🔍 Read the full analysis: AI Shopping Agents Put Retailer Control In The Spotlight on ThorstenMeyerAI.com
Get smart everyday buys delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Meta says more than 10 million people in the United States and Canada set up its Muse shopping agent within about six weeks, but the agent cannot shop across every retailer. Walmart is building an access route for approved agents, while Amazon has blocked Muse and says it violated the retailer’s terms. The competing approaches leave questions about shopper control, payments, privacy and how publishers receive credit for purchases.
Walmart is opening a structured route for approved AI shopping agents while Amazon has blocked Meta’s Muse, sharpening a dispute over who controls purchases made by software on shoppers’ behalf. Meta said on October 21 that more than 10 million people in the United States and Canada had set up Muse in roughly six weeks, but the agent’s access depends on each retailer’s rules.
Walmart announced Agent Profiles and an Agent Gateway on October 21. The gateway is designed to let authorized agents search inventory and make purchases through a defined technical interface, rather than navigating a store’s consumer website as a person would. Agent Profiles are intended to set permissions, including spending limits. The announcement makes Walmart one of the retailers explicitly developing a way for outside agents to interact with its shopping services.
Meta’s figures come from its third-quarter earnings call, not an independent measurement. The company said shopping is the most common use of Muse and that the average purchase is more than $200. The source does not specify the period or calculation behind that average. A Verge reporter’s test illustrated the limits of the service: Muse named Sam’s Club, Walmart and Costco as options for granola bars, then said Amazon might have a lower price but it could not check because Amazon had blocked it.
Amazon said Muse’s access violated its Conditions of Use. The retailer’s stated concerns included that Meta had not disclosed the agent’s shopping activity, that the agent did not identify itself, and that it appeared to capture and store customer credentials. Meta disputed the password and payment concern in its launch materials, saying Muse had no visibility into users’ passwords or payment methods. Those are competing descriptions of the agent’s handling of sensitive information, not a settled finding about what happened.
Agentic shopping
Three ways an AI agent buys for you
Every fight between shops and AI agents comes down to which of these three routes the agent uses, and whether the shop agreed to it.
The browser, like a human
The agent opens pages built for people, signs in with your login and clicks through checkout. Works almost anywhere, which is why shops object: they can't tell the agent from you.
Examples: Meta's Muse (from a virtual machine in Meta's cloud), Perplexity's Comet (from your own computer)
The shop's own agent door
Approved agents search and buy through a structured interface instead of clicking through pages. The shop sets the rules.
Examples: Walmart Agent Gateway (MCP, beta), Shopify MCP server and WebMCP checkout tools
Shared protocols
Common rules for who the agent is, what you allowed, and how the checkout and payment are confirmed.
Examples: Personal Agent Protocol (draft), Universal Commerce Protocol, Agentic Commerce Protocol
The protocols competing to be the standard
| Protocol | Backers | Covers | Status |
|---|---|---|---|
| Personal Agent Protocol (Poppy) | Sierra, Meta, Walmart, Shopify, Stripe + 35 design partners | Agent identity, customer permission, one session across site, API and company agent | Draft 0.1 (9 Oct 2026) |
| Universal Commerce Protocol (UCP) | Google, Shopify; site also lists Etsy, Wayfair, Target, Walmart, Amazon, Microsoft | Discovery, cart, checkout, orders; payment mandates via AP2 | Published |
| Agentic Commerce Protocol (ACP) | OpenAI, Stripe | Checkout between agent and merchant | Open source; ChatGPT's own checkout scaled back in March 2026 |
What a careful agent purchase looks like
- You give the task and a limit, for example a $50 cap.
- The agent identifies itself to the shop and gets only the access you approved.
- It prepares the cart and shows you an approval card.
- Payment goes through a token or one-time card, so the agent never sees your card.
- The shop’s normal returns apply if the agent got it wrong.
Sources: Meta (8 Sep 2026), Walmart (21 Oct 2026), TechCrunch on Shopify (28 Sep 2026), Sierra (9 Oct 2026), ucp.dev, agenticcommerce.dev, OpenAI (24 Mar 2026). Purchase steps: Walmart’s agent terms and Meta’s Muse description.
Powered by Thorsten Meyer AI
Who Controls the Agent Checkout
The dispute concerns more than whether an agent can find a cheaper item. The company controlling the agent may shape which shops appear, how products are compared, where a purchase is completed and which payment or return process applies. Retailers, in turn, have reasons to control automated access to their inventory, customer accounts and checkout systems. Access rules can determine which products an agent can compare and whether a retailer can maintain its own relationship with the buyer.
The source points to a potential effect on advertising and referrals. It says Amazon generated more than $68 billion from advertising last year, citing GeekWire, while Meta has said it expects to make money by taking a small fee from transactions. Those figures and statements describe different business models; they do not establish how much either company will earn from agent shopping. Publishers and affiliate sites may also lose referral credit when an agent reads a review but completes a purchase without using a tracked link. Amazon’s Associates terms, as described in the source, tie commissions to clicks on such links.
Early traffic data suggests many agents are still researching rather than completing transactions. HUMAN Security’s September data, as cited in the source, found that 81% of agent requests went to product and search pages, while 2.4% went to checkout and payment. The figures describe agent requests in that dataset, not all shopping activity or all agent users. Still, they indicate that the immediate contest may be about product discovery and access as much as payment.
Products Worth Considering
Samsung 27-inch Odyssey OLED G6 QD-OLED QHD Gaming Monitor
Dell UltraSharp 27 Thunderbolt Hub Monitor (U2726DER)
Samsung 27-Inch Odyssey QD-OLED G8 4K 240Hz Gaming Monitor
As an affiliate, we earn on qualifying purchases.
From Browser Bots to Retail APIs
Meta launched Muse on September 8, describing an agent that can open a browser, fill out forms and negotiate for users. It runs in a dedicated virtual machine in Meta’s cloud and uses Link by Stripe, which issues a one-time-use card, according to the source. Amazon blocked Muse on September 20. Three days later, Meta announced retail connections with Best Buy, Gap, Sephora, Walmart, Wayfair and Expedia.
Retailers and technology companies are trying several ways to make agent shopping work. A browser-based agent acts through pages designed for people; a retailer’s own interface can instead grant an agent controlled access. Walmart’s gateway uses Model Context Protocol (MCP), while Shopify has offered structured tools for browser agents to handle checkout. Industry groups are also developing protocols to identify agents and pass along customer permissions. These efforts remain at different stages: the Personal Agent Protocol, announced by Sierra and Meta with partners including Walmart, Shopify and Stripe, was a draft in October, while other commerce protocols have been published.
Earlier efforts show that building a checkout connection does not guarantee that it will become a standard. OpenAI introduced Instant Checkout in ChatGPT using the Agentic Commerce Protocol in September 2025. In March 2026, it said the first version lacked the flexibility it wanted and that it would let merchants use their own checkout experiences while OpenAI focused on product discovery. That retreat underlines how difficult it can be to align agents, retailers and payment systems.
“Muse has no visibility into people’s passwords or payment methods.”
— Meta, in its Muse launch materials
Access, Safety and Purchase Rules
It is not yet clear how many Muse users complete purchases, how its reported average purchase value was calculated, or how much shopping activity will move through retailer-approved agent connections. Meta’s usage figures are company-reported, and the HUMAN Security data covers requests in its own September dataset. Neither establishes the share of all online shopping conducted by agents.
The details of the Amazon-Meta dispute also remain contested. The supplied material reports Amazon’s concerns and Meta’s denial about access to passwords and payment methods, but does not independently resolve them. It is also unclear which retailers will allow browser-based agents, whether shoppers will be able to compare prices across blocked stores, and how refunds, returns and disputes will be handled when an agent places an order. Protocols may help identify an agent and record permissions, but their adoption and compatibility are still developing.
Retailer Policies Will Set the Path
The next practical test is whether Walmart’s gateway and spending profiles work reliably for shoppers and approved agents, and whether other retailers adopt similar access rules. The protocol partners will also need to turn draft specifications into systems that merchants, agents and payment providers can use together. Retailer decisions will shape where agents can shop before any one technical standard can establish broad access.
For shoppers, the immediate safeguards are straightforward: check whether a purchase requires approval, set a spending cap, understand how payment details are handled and confirm which seller’s return policy applies. The source says Muse uses a one-time-use card through Link by Stripe, but that does not answer every question about order changes, refunds or customer support. Until those processes are clear for each service, convenience should not be mistaken for a uniform shopping experience.
Key Questions
What is the main development in AI shopping?
Walmart announced an Agent Gateway and shopper profiles with permissions, while Amazon has blocked Meta’s Muse. The contrast shows that agents’ access to stores is being decided retailer by retailer.
Why did Amazon block Meta’s Muse?
Amazon said Muse’s access violated its Conditions of Use, citing concerns about disclosure, agent identification and possible credential handling. Meta disputed the password and payment concern, saying Muse could not see those details; the supplied material does not independently resolve the disagreement.
How many people have set up Muse?
Meta said more than 10 million people in the United States and Canada set up Muse within about six weeks. That is a company-reported figure from its October 21 earnings call.
Can AI shopping agents complete purchases?
Some are designed to do so, but access varies. Agents may use a browser, a retailer’s approved interface or a shared protocol. Retailer policies and checkout integrations determine whether an agent can complete a purchase at a particular store.
What should shoppers check before using an agent?
Review purchase-approval settings and spending limits, and check how payment details, returns and refunds are handled. The applicable return policy may depend on the retailer where the order is placed.
Source: ThorstenMeyerAI.com
